又一重磅折腾，在PVE8中使用CT容器Debian安装comfyui绘画工具及NVIDIA Container Toolkit安装

前言

折腾waydroid时，重装了comfyui所在的宿主机系统。想了一下，为了更方便的折腾，把这个电脑也做成了虚拟化平台！抄作业过程如遇到问题，自行补充。因子方放的内容，不包括之前错误的，可能有些看似错误的步骤，对本次成功安装是有帮助的。

nvdia驱动卸载

可能会一直搞坏驱动，更换版本之类的，放一些卸载的命令，结合着用吧。例如使用run文件安装的驱动，应该使用nvidia-uninstall命令，而非apt。此外说一点，之所以使用run文件安装，是因为apt 安装时，会触动pve的钩子文件，导致无法正常安装。

nvidia-uninstall
rm -rf /usr/local/cuda*
rm -rf /etc/modprobe.d/nvidia*
rm -rf /var/lib/nvidia
rm -rf /usr/share/nvidia
rm -rf /usr/bin/nvidia*
rm -rf /usr/lib/x86_64-linux-gnu/nvidia

sudo nvidia-uninstall
sudo apt purge 'nvidia-*' -y
sudo rm -rf /usr/local/cuda*
sudo rm -f /usr/lib/x86_64-linux-gnu/libcuda*
sudo rm -f /usr/lib/x86_64-linux-gnu/libnvidia*
sudo rm -rf /etc/modprobe.d/nvidia*
sudo rm -rf /lib/modules/$(uname -r)/kernel/drivers/video/nvidia*

自动配置

apt install pve-nvidia-vgpu-helper
pve-nvidia-vgpu-helper setup

pve-nvidia-vgpu-helper工具将设置一些基本功能，例如将nouveau驱动程序列入黑名单、安装头文件包、DKMS 等等。

确认一些信息

如果没有安装驱动，下面只有一个crad和render。要记住哪个没有，安装驱动后出现的，那个就是nvdia的显卡，后面我们要绑定到lxc里面的。（当然没记住，后面再找就麻烦一点，这个是比较简单的办法）。我一开始就是没绑定card和render，无法进行计算。
另外补充一点，宿主机重新开机，启动CT提示/dev/nvidia-uvm没有挂载的，可以在宿主机允许nvidia-smi，再开启CT

root@mini:~# ls -l /dev/dri
total 0
drwxr-xr-x 2 root root        120 Jul 31 14:16 by-path
crw-rw---- 1 root video  226,   0 Jul 31 14:16 card0
crw-rw---- 1 root video  226,   1 Jul 31 14:16 card1
crw-rw---- 1 root render 226, 128 Jul 31 14:16 renderD128
crw-rw---- 1 root render 226, 129 Jul 31 14:16 renderD129

驱动下载

指引页面：https://www.nvidia.com/en-us/drivers/unix/
全部驱动：https://download.nvidia.com/XFree86/Linux-x86_64/

宿主机安装驱动

安装

考虑到显卡老爷爷，选择535
wget https://download.nvidia.com/XFree86/Linux-x86_64/535.261.03/NVIDIA-Linux-x86_64-535.261.03.run
sh NVIDIA-Linux-x86_64-535.261.03.run

验证1，确认驱动被正确识别

root@mini:~# nvidia-smi
Thu Jul 31 11:02:15 2025       
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.261.03             Driver Version: 535.261.03   CUDA Version: 12.2     |
|-----------------------------------------+----------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |         Memory-Usage | GPU-Util  Compute M. |
|                                         |                      |               MIG M. |
|=========================================+======================+======================|
|   0  NVIDIA P104-100                Off | 00000000:01:00.0 Off |                  N/A |
| 63%   42C    P0              42W / 180W |      0MiB /  8192MiB |      0%      Default |
|                                         |                      |                  N/A |
+-----------------------------------------+----------------------+----------------------+
                                                                                         
+---------------------------------------------------------------------------------------+
| Processes:                                                                            |
|  GPU   GI   CI        PID   Type   Process name                            GPU Memory |
|        ID   ID                                                             Usage      |
|=======================================================================================|
|  No running processes found                                                           |
+---------------------------------------------------------------------------------------+

验证2，找到需要绑定的设备

root@mini:~# ls -al /dev/nvidia*
crw-rw-rw- 1 root root 195, 0 Jul 31 10:24 /dev/nvidia0
crw-rw-rw- 1 root root 195, 255 Jul 31 10:24 /dev/nvidiactl
crw-rw-rw- 1 root root 510, 0 Jul 31 10:57 /dev/nvidia-uvm
crw-rw-rw- 1 root root 510, 1 Jul 31 10:57 /dev/nvidia-uvm-tools

/dev/nvidia-caps:
total 0
drwxr-xr-x 2 root root 80 Jul 31 10:57 .
drwxr-xr-x 19 root root 4280 Jul 31 11:02 ..
cr-------- 1 root root 235, 1 Jul 31 10:57 nvidia-cap1
cr--r--r-- 1 root root 235, 2 Jul 31 10:57 nvidia-cap2

# 如果是桌面环境，上面还应该显示/dev/nvidia-modeset

设备绑定

在对应lxc.conf中添加以下内容，或者在web界面绑定

dev0: /dev/nvidia0
dev1: /dev/nvidiactl
dev2: /dev/nvidia-uvm
dev3: /dev/nvidia-uvm-tools
dev4: /dev/nvidia-caps/nvidia-cap1
dev5: /dev/nvidia-caps/nvidia-cap2
dev6: /dev/dri/card0,gid=44
dev7: /dev/dri/renderD129,gid=104

推送到CT容器安装

pct push 100 NVIDIA-Linux-x86_64-535.261.03.run /root/NVIDIA-Linux-x86_64-535.261.03.run
sh NVIDIA-Linux-x86_64-535.261.03.run --no-kernel-modules

# LXC 容器共享宿主机的内核，所以安装时需要指定 –no-kernel-module 参数，避免安装内核模块

容器验证1，确认驱动正常识别

root@Comfy:~# nvidia-smi
Thu Jul 31 11:13:16 2025
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA P104-100 Off | 00000000:01:00.0 Off | N/A |
| 64% 42C P0 39W / 180W | 0MiB / 8192MiB | 0% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+

+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| No running processes found |
+---------------------------------------------------------------------------------------+

容器验证2，绑定状态是否一致

root@Comfy:~# ls -al /dev/nvidia*
crw-rw---- 1 root root 509,   0 Jul 31 10:36 /dev/nvidia-uvm
crw-rw---- 1 root root 509,   1 Jul 31 10:36 /dev/nvidia-uvm-tools
crw-rw-rw- 1 root root 195,   0 Jul 31 10:36 /dev/nvidia0
crw-rw-rw- 1 root root 195, 255 Jul 31 10:36 /dev/nvidiactl

/dev/nvidia-caps:
total 0
drwxr-xr-x 2 root root     80 Jul 31 10:36 .
drwxr-xr-x 7 root root    580 Jul 31 10:36 ..
crw-rw---- 1 root root 234, 1 Jul 31 10:36 nvidia-cap1
crw-rw---- 1 root root 234, 2 Jul 31 10:36 nvidia-cap2

常规安装操作

apt install sudo git
sudo usermod -aG sudo l
apt install python3-venv python3-full -y

# 验证下版本，等下要使用对应的wheel文件。详情看之前的文章
l@Comfy:~$ python3 --version
Python 3.11.2

准备虚拟环境

python3 -m venv ~/comfyui-env
source ~/comfyui-env/bin/activate
pip install --upgrade pip
pip config set global.index-url https://mirrors.cernet.edu.cn/pypi/web/simple

手动安装准备的文件

pip install /home/l/torch-2.6.0+cu118-cp311-cp311-linux_x86_64.whl
pip install /home/l/torchaudio-2.6.0+cu118-cp311-cp311-linux_x86_64.whl
pip install /home/l/torchvision-0.21.0+cu118-cp311-cp311-linux_x86_64.whl

下载对应源码

git clone --depth=1 https://github.com/comfyanonymous/ComfyUI
cd /home/l/ComfyUI/
pip install -r requirements.txt

??完工！！
python main.py --listen 0.0.0.0

??一键进入命令
source ~/comfyui-env/bin/activate && cd ComfyUI && python main.py --listen 0.0.0.0

NVIDIA Container Toolkit

本来以为要成功了，结果跑图就崩溃了。又继续研究了一番，主要是添加了NVIDIA Container Toolkit。先放官方链接指引：https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html

apt install gpg curl

curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
&& curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

sudo apt-get update

export NVIDIA_CONTAINER_TOOLKIT_VERSION=1.17.8-1
sudo apt-get install -y \
nvidia-container-toolkit=${NVIDIA_CONTAINER_TOOLKIT_VERSION} \
nvidia-container-toolkit-base=${NVIDIA_CONTAINER_TOOLKIT_VERSION} \
libnvidia-container-tools=${NVIDIA_CONTAINER_TOOLKIT_VERSION} \
libnvidia-container1=${NVIDIA_CONTAINER_TOOLKIT_VERSION}

再次运行，正常了！

喜欢 (0)赏

[赞赏使我快乐]

分享 (0)